Pseudo R² for meqrlogit using mi estimate

[Yvonne Krieg]

Hey,

I found the following code in this forum to get the pseudo R² for imputed data:

local rhs "armg2 armg3 tbsaburn20 tbsaburn21"
noi mi estimate, or saving(miest, replace): logistic hodc `rhs', vce(cluster site)
qui mi query
local M=r(M)
scalar r2=0
scalar cstat=0
qui mi xeq 1/`M': logistic hodc `rhs'; scalar r2=r2+e(r2_p); lroc, nog; scalar cstat=cstat+r(area)
scalar r2=r2/`M'
scalar cstat=cstat/`M'
noi di "Pseudo R=squared over imputed data = " r2
noi di "C statistic over imputed data = " cstat

I have adapted the code for my needs:

local rhs "geschl_n alter_n ethnieend2 schform sozhilf_n selcon bsiagg emp hiesel sozdep eltgew herumf rexnied rexmut rexvat rexle rexfr reakmut_pos reakvat_pos reakle_pos reakfr_pos aus_mit"
noi mi estimate, or saving(miest, replace): meqrlogit taetaus `rhs', or || klcode: ,
qui mi query
local M=r(M)
scalar r2=0
scalar cstat=0
qui mi xeq 1/`M': meqrlogit taetaus `rhs', or || klcode: ,; scalar r2=r2+e(r2_p); lroc, nog; scalar cstat=cstat+r(area)
scalar r2=r2/`M'
scalar cstat=cstat/`M'
noi di "Pseudo R=squared over imputed data = " r2
noi di "C statistic over imputed data = " cstat

However, after the "qui mi xeq" command I get the error message "Invalid numlist r(121);". Do you have any idea what I could have done wrong?

Thanks in advance,
Yvonne

[Carlo Lazzaro]

Yvonne:
welcome to this forum.
What if you call the local in your code as:

"`M'"

instead of;

`M'

[Yvonne Krieg]

Thank you for your suggestion. Unfortunately, I get the same error message, if I understood your suggestion correctly.

local rhs "geschl_n alter_n ethnieend2 schform sozhilf_n selcon bsiagg emp hiesel sozdep eltgew herumf rexnied rexmut rexvat rexle rexfr reakmut_pos reakvat_pos reakle_pos reakfr_pos aus_mit"
noi mi estimate, or saving(miest, replace): meqrlogit taetaus `rhs', or || klcode: ,
qui mi query
local M=r(M)
scalar r2=0
scalar cstat=0
qui mi xeq 1/"`M'": meqrlogit taetaus `rhs', or || klcode: ,; scalar r2=r2+e(r2_p); lroc, nog; scalar cstat=cstat+r(area)
scalar r2=r2/`M'
scalar cstat=cstat/`M'
noi di "Pseudo R=squared over imputed data = " r2
noi di "C statistic over imputed data = " cstat

[Carlo Lazzaro]

Yvonne:
try your original code (ie, the one you originally posted) just omitting the comma after -klcode-.

[Yvonne Krieg]

Unfortunately the same error message

qui mi xeq 1/`M': meqrlogit taetaus `rhs', or || klcode:; scalar r2=r2+e(r2_p); lroc, nog; scalar cstat=cstat+r(area)

[Carlo Lazzaro]

Yvonne:
the usual recipe is to go on with code lines step by step tocheck where Stata throws the error message.
If the issue is in the line strating with -qui mi xeq 1- stop, for instance, at -klcode:- and see what happens.
This approach, that heavily draws upon the -what to do when one engine flames out- pilots checklist, sometimes is successful in sniffing out the culprit .

[Rich Goldstein]

I think you need to set trace on (with a depth of 1) so we can see exactly where the problem is; see

help trace

[Yvonne Krieg]

Thank you for encouraging me to check the code step by step. I think it has something to do with the ", or" option as I got a different error message after I deleted it. Unfortunately it still did not work.

I have now used a workaround by running a "normal" logistic regression with clustred standard errors instead of a multilevel model, since I have no second level variables.
This worked:

local rhs "i.geschl_n alter_n ethnieend2 i.schform i.sozhilf_n selcon bsiagg emp hiesel sozdep eltgew herumf rexnied rexmut rexvat rexle rexfr i.reakmut_pos i.reakvat_pos i.reakle_pos i.reakfr_pos aus_mit"
noi mi estimate, or saving(miest, replace): logit taetaus `rhs', vce(cluster klcode)
qui mi query
local M=r(M)
scalar r2=0
scalar cstat=0
qui mi xeq 1/`M': logit taetaus `rhs'; scalar r2=r2+e(r2_p); lroc, nog; scalar cstat=cstat+r(area)
scalar r2=r2/`M'
scalar cstat=cstat/`M'
noi di "Pseudo R=squared over imputed data = " r2
noi di "C statistic over imputed data = " cstat

Thank you again!